DescriptionThis thesis presents an adaptive MAC (AMAC) protocol for supporting MAC layer adaptation in cognitive radio networks. MAC protocol adaptation is motivated by the exibility of emerging software-de ned radios which make it feasible to dynamically adjust radio protocols and parameters in order to maintain communications quality. Dynamic changes to the MAC layer may be useful in tactical or vehicular networking scenarios, where radio node density, tra c volumes and service requirements can vary widely over time. A speci c control framework for the proposed AMAC algorithm is described based on the "CogNet" protocol stack which uses a Global Control Plane (GCP) to distribute control information between nearby radios. An AMAC prototype which switches between CSMA and TDMA is evaluated for various tra c scenarios using the NS-2 simulator. In addition, a proof-of-concept AMAC is implemented using GNUradio/USRP platforms on the ORBIT radio grid testbed. Detailed simulation and experimental results are given for both UDP and TCP tra c with di erent usage scenarios and application models. The results show that AMAC can provide improved performance relative to a conventional static system and can be implemented with reasonable control protocol overhead and latency.